SELECTED PROJECTGenerative AI

Cascaded Text-to-Image Diffusion Model

An Imagen-inspired diffusion cascade separating text-conditioned generation from super-resolution.

1024²OUTPUT RESOLUTION
  • Python
  • PyTorch
  • Diffusion

Objective

Build a text-to-image system that increases output resolution through separate generative stages.

System

Implemented a multi-stage diffusion pipeline inspired by Imagen with Python and PyTorch, separating text-conditioned image generation from super-resolution.

Outcome

The cascade progressively produces images at resolutions up to 1024 × 1024. No formal image-quality score is claimed.